AI instruction set is logged on start as null

[artifishvr]

Likely a regression in #78? Can't exactly find it but the production instance of pathways hasn't been updated to v3 and doesn't have this issue.

[daniwasonline]

What instruction set do you have set in your environment? I can't repro this issue with MODEL_LLM_PRESET=spongeass (3.0.0-0.alpha).

[artifishvr]

MODEL_LLM_PRESET=spongeass

[artifishvr]

my entire env file is (with sensitive keys made xxxxxxxx)

# !! Core Discord application settings

# Discord bot token; requires MESSAGE_CONTENT intent, see: [link to be added later]
DISCORD_TOKEN=xxxxxxxx
APPLICATION_ID=xxxxxxxx

# !! Environment context

# Specify the Node environment context that SpongeChat should be running in.
## Usually, env#NODE_ENV is either one of PRODUCTION, DEVELOPMENT, or STAGING.
NODE_ENV=DEVELOPMENT

# These keys are used for developer mode and debugging.
DEV_GUILD=xxxxxxxx
DEV_UNLOCK_ROLE=

# !! Database (k/V)

# A Valkey (Redis OSS)-compatible key/value store is required to run SpongeChat 2.
## Although Dragonfly uses the BSL, it is currently the only Redis-compatible k/V store that supports JSON.
## Dragonfly has decent corporate goodwill and the BSL changes to OSS after 4 years. As such, we use it.
KV_URI=redis://127.0.0.1:6379

# !! Model activation settings

# Select a mode for model activation.
## WHITELIST only allows channels specified in ACTIVATION_LIST.
## BLACKLIST allows all channels other than those specified in ACTIVATION_LIST.
ACTIVATION_MODE=WHITELIST

# A comma-separated list of channel IDs. Read env#ACTIVATION_MODE for more information.
ACTIVATION_CHANNEL_SETLIST=1229595792885682176

# !! Model settings

# Workers AI: Specify a Cloudflare account ID and token for use with Workers AI.
## SAFETY WARNING: Scope your API token to only be accessible with Workers AI.
CLOUDFLARE_ACCOUNT_ID=xxxxxxxx

CLOUDFLARE_ACCOUNT_TOKEN=xxxxxxxx

# The following variables specify instruction sets and configuration for the various models in SpongeChat.
## The value of the variable should correspond to the key of its' responding configuration in modelConstants.js.
MODEL_LLM_PRESET=spongeass

## Callsystems are used to call functions during passive activation.
### Legacy is the older function system. While it only supports image generation, it does not require additional models other than the base text model. This is also the fallback system used when Integrations fails.
### To use Legacy, specify "cs.spongedsc.legacy".
MODEL_LLM_CALLSYSTEM=cs.spongedsc.legacy
## If callsystem versioning is enabled, specify the version of the callsystem to use. Use `latest` to use the latest version.
MODEL_LLM_CALLSYSTEM_VERSION=latest

# !! Memory-viewer (pathways-web)

WEB_HOST=https://remember.gayse.cx
WEB_KEY=xxxxxxxx

[daniwasonline]

what in the name of crack